On Sun, Mar 16, 2025 at 01:43:49PM +0800, Herbert Xu wrote: > On Sat, Mar 15, 2025 at 09:49:37PM -0700, Eric Biggers wrote: > > > > As I've said before, this would be much better handled by a function that > > explicitly takes multiple buffers, rather than changing the whole API to make > > every request ambiguously actually be a whole list of requests (whose behavior > > also differs from submitting them individually in undocumented ways). > > This is exactly how we handle GSO in the network stack. It's > always an sk_buff regardless of whether it's a batch or a single > one. We don't have to make the same mistakes again. > In fact I will be using that to handle GSO over IPsec so the > array-based interface that you proposed simply does not fit. IPsec doesn't use compression, so I guess your comment above is about hashing. But you are *still* ignoring all the reasons why it isn't useful for IPsec. Nacked-by: Eric Biggers <ebiggers@xxxxxxxxxx> - Eric
